Throughout the 19th and 20th centuries, scientists and physicians dreamt of being able to custom design therapeutic agents that were highly specific for a single biological target. By selectively attacking disease while sparing healthy tissue, these “magic bullets” were thought to be ideal therapeutic agents. It was not until the early 1970's, however, that this dream began to be realized when Kohler and Milstein (Nature 256: 495, 1975) developed a ground-breaking method for making antigen (Ag)-specific monoclonal antibodies (mAbs). This was perhaps one of the most significant events in the history in modern immunology. Using this new approach, numerous research, diagnostic, and therapeutic products were developed with a combined market of hundreds of billions of dollars.
In Kohler and Milstein's method, rodent B cells which each express a unique immunoglobulin (Ig) (i.e., a mAb) were immortalized by fusion with a myeloma cell. The ability of these mAb-secreting “hybridomas” to replicate ad infinitum allowed those that produced a mAb with high affinity for a given Ag to be selected as an engine for producing large quantities of identical mAbs. MAbs made by this method eventually were used as human therapeutics. While effective, the use of therapeutic rodent mAbs was soon discovered to be less than ideal because, after multiple administrations, they induced anti-rodent Ig immune responses, which neutralized the activity of the mAb and frequently caused serious adverse reactions in treated patients.
Attempts were therefore made to adapt Kohler and Milstein's rodent method to make human mAbs by fusing human B cells to myeloma cells. This “human hybridoma” technique proved to be more challenging than expected for a number of different reasons. First, for ethical or medical reasons, it was often difficult to increase the frequency of Ag-specific B cells by repeatedly immunizing human subjects with an Ag. Second, also for ethical or medical reasons, it was not practical to isolate human spleen tissue to obtain B cells as was done in rodents. Third, immortalized mAb-producing human cell lines were not easy to establish and tended to produce very little mAb. And fourth, because the human immunoglobulin diversity, or repertoire, of individual Ab specificities is much higher than the rodent repertoire (˜1012 vs. ˜107 individual specificities), and because the screening methodology employed was limited to no more than about 104 specificities in a given sampling, it is much more technically difficult to isolate human mAb-secreting cells of a desired specificity. Identifying mAbs having low frequency specificities by this technique thus remains a daunting, “finding a needle in the haystack” effort.
In parallel with the development of “human hybridoma” methods, scientists also began attempting to “humanize” rodent antibodies by replacing various rodent amino acid sequences with those from human antibodies. In a typical application, murine Ig variable regions are combined with human constant domains, or murine complementarity determining regions (CDRs) are grafted onto a human Ig framework. This new recombinant nucleic acid construct is inserted into an expression vector which is transfected into host cells that can produce large quantities of the chimeric mAb when cultured. While this technique has done much to improve the usefulness of mAbs in therapeutic applications (by reducing the human anti-rodent-immunoglobulin response), the remaining rodent sequences can still cause the development of an undesirable anti-mAb response [e.g., a HAMA (human anti-mouse antibody) response] that can neutralize the mAb or even cause a serious adverse reaction in a patient.
Another approach for producing human mAbs has been to use chimeric animals wherein the host animal's Ig genes have been replaced, in part, with their human counterpart. Upon vaccination, these knockout/knockin animals produce human Abs and their B cell-containing spleens can be used to make human mAbs using the conventional hybridoma technique. Shortcomings that have been associated with this technology include: reduced diversity (specificity) and affinity due to an incomplete repertoire of human Ig-related genes; humoral responses that are not purely human (“humanized” transgenic animals retain their native T cells and have “leaky” expression of murine antibody responses); the antibodies produced have non-human glycosylation patterns (e.g., Galα 1-3Gal containing oligosacarides in the humanized animals are not commonly found in humans and are thus recognized as foreign); and antibodies that undergo development in an animal are not “selected” for tolerance in a human—thus even though derived from partial human sequences, they are not necessarily non-immunogenic in humans (i.e., not true human).
The advent of polymerase chain reaction (PCR) methods made it possible to amplify entire or parts of Ig genes, and allowed the creation of libraries of plasmids containing cDNAs encoding heavy and light Ig chain variable regions (V-regions) from peripheral blood leukocytes (PBLs) to be produced. Traditionally, library generation for phage display involved a set of 6 primers in the variable region of heavy chain, 4 primers in the variable region of kappa chain, and 9 primers in the variable region of lambda chain. (as described in Barbas et al., Phage Display: A Laboratory Manual. Cold Spring Harbor, 2001.) The plasmids from these libraries can be used to transfect bacteria or other host cells (e.g., yeast) which can then be screened for Ag-specific Ig components. In the phage display approach, human Ig V-region genes are cloned into bacteriophage in order to display Ab fragments on the surfaces of bacteriophage particles. Genes from extensive human Ab gene libraries are inserted into a library of phage. Each phage can potentially carry the gene for a different V-region and displays a different V-region on its surface. Commonly these phage are used to display on a bacteriophage surface an Ab construct made up of heavy and light Ig chain V-regions separated by a linker (a single chain fragment variable Ab or scFv), such that each heavy and light chain V-regions can bind a target antigen in a manner that closely mimics how a mAb binds Ag. Similar to phage display, in yeast display, Ig components are displayed as a fusion to the Aga2p protein on the surface of yeast in a manner that projects the fusion protein away from the cell surface. Flow cytometry can then be used to select those yeast cells expressing an Ig component that reacts with a target Ag. Once identified by phage or yeast display, the expressed heavy and/or light chain V-regions genes are recovered and inserted into a full-length heavy and/or light chain expression vectors. Culturing host cells transfected with such vectors allows large quantities of a mAb to be produced.
The primary benefit of phage or yeast display techniques is that huge libraries, incorporating as many as 100 billion distinct V-regions, may be screened for Ag reactivity. A drawback of these techniques is that the V-regions isolated in the first round of display are typically of very low affinity. A common reason for generating low affinities antibodies is the result of the loss of crucial hypermutated sequences (which create the mature high affinity antibodies) using this method. B lymphocytes that produce high affinity antibodies have generally undergone a process known as somatic hypermutation. In this process, germline encoded V-region sequences undergo genetic mutations, such that the gene sequence of the V-region is altered. This can result in the production of a “mature,” V-region sequence that has substantially higher-affinity for a given antigen. It is generally understood that somatic hypermutation is a fundamentally important event for generating antibody diversity and high affinity antibodies. This somatic hypermutation cannot be found in the germline genomic sequences, and is represented only in a unique B lymphocyte clone.
The generation of low affinity antibodies using the phage display approach has been a major limitation to this method. This difficulty has indeed resulted in the development of many other complex approaches, including humanized mice and in vitro somatic hypermutation, as described above. Antibodies in humans undergo a selection process that makes them tolerated in the human body. These techniques result in the use of antibody sequences that are not based on human in vivo selection processes, and thus these antibodies can be immunogenic in humans. In any event, overcoming the production of low-affinity antibodies has been a major challenge for these methods.
A reason why phage display commonly generates low affinity antibodies is that it involves the use of primer sequences that are based on germline V-region sequences. There exist two possible outcomes from using V-region primers based on germline sequences: (1) the primers will fail to hybridize with somatically mutated sequences, since the mismatch between the primer and hypermutated sequences fails to result in primer annealing; or, (2) the mismatch between the primer and hypermutated sequences is conservative, such that annealing does occur, but the amplification results in recapitulation of the germline-sequence encoded by the primer rather than the underlying hypermutated sequence. In either event, since the phage display methodology involves crucial PCR amplification steps using PCR primer sequences that encode germline V-region sequences, and since these primers amplify the Ig sequences from inside the antibody V-regions, cDNA libraries generated using germline primer sequences will not contain the full repertoire of somatic hypermutated sequences. Furthermore, and importantly, the repertoire will be completely devoid of antibodies with hypermutation in the 5′ end of the V-region, i.e., the region encoded by the germline encoded primer.
In other words, using primers based on germline sequences recapitulates the germline V-region sequences, thereby losing hypermutations in the area encoded by the primer sequence. This method thus systematically eliminates hypermutation in these areas of the V-region and results in an overall loss of hypermutated antibody sequences. In the case of low frequency Abs (i.e., less than 0.1, 0.01, 0.001, 0.0001, or 0.00001% of a subject's entire repertoire of Ag specificities), for which the phage display system is arguably most suited, finding V-region genes that can be used to make a high affinity mAb becomes all the more challenging.